Fix the failing template test for AVX10.2 (#112252)

* Limit JIT unit suite within the subsets which are stable in SDE.

* Fix the bug in instruction definition for INS_VMPSADBW

* Revert "Limit JIT unit suite within the subsets which are stable in SDE."

This reverts commit 552bbe2.

---------

Co-authored-by: Ruihan-Yin <ruihan.yin@intel.com>

src/coreclr/jit/instrsxarch.h

@@ -915,7 +915,7 @@ INST3(vcvtps2ibs,       "cvtps2ibs",        IUM_WR, BAD_CODE,               BAD_
 INST3(vcvtps2iubs,      "cvtps2iubs",       IUM_WR, BAD_CODE,               BAD_CODE,     PCKDBLMAP(0x05, 0x6B),       INS_TT_FULL,                         Input_32Bit    | REX_W0       | Encoding_EVEX)                                                                                                                                  // cvt with truncation/saturation scalar single to unsigned DWORD/QWORD
 INST3(vcvttps2ibs,      "cvttps2ibs",       IUM_WR, BAD_CODE,               BAD_CODE,     PCKDBLMAP(0x05, 0x68),       INS_TT_FULL,                         Input_32Bit    | REX_W0       | Encoding_EVEX)                                                                                                                                  // cvt with truncation/saturation scalar single to unsigned DWORD/QWORD
 INST3(vcvttps2iubs,     "cvttps2iubs",      IUM_WR, BAD_CODE,               BAD_CODE,     PCKDBLMAP(0x05, 0x6A),       INS_TT_FULL,                         Input_32Bit    | REX_W0       | Encoding_EVEX)                                                                                                                                  // cvt with truncation/saturation scalar single to unsigned DWORD/QWORD
-INST3(vmpsadbw,         "mpsadbw",          IUM_WR, BAD_CODE,               BAD_CODE,     AVX3A(0x42),                 INS_TT_FULL,                         Input_8Bit     | REX_W0       | Encoding_EVEX             | INS_Flags_IsDstDstSrcAVXInstruction)                                                                                           // Compute Multiple Packed Sums of Absolute Difference
+INST3(vmpsadbw,         "mpsadbw",          IUM_WR, BAD_CODE,               BAD_CODE,     AVX3A(0x42),                 INS_TT_FULL_MEM,                     Input_8Bit     | REX_W0       | Encoding_EVEX             | INS_Flags_IsDstDstSrcAVXInstruction)                                                                                           // Compute Multiple Packed Sums of Absolute Difference
 
 INST3(vminmaxsd,        "minmaxsd",         IUM_WR, BAD_CODE,               BAD_CODE,     SSE3A(0x53),                 INS_TT_TUPLE1_SCALAR,                Input_64Bit    | REX_W1       | Encoding_EVEX             | INS_Flags_IsDstSrcSrcAVXInstruction)                                                                                                                                  // Return Minimum/Maximum scalar double 
 INST3(vminmaxss,        "minmaxss",         IUM_WR, BAD_CODE,               BAD_CODE,     SSE3A(0x53),                 INS_TT_TUPLE1_SCALAR,                Input_32Bit    | REX_W0       | Encoding_EVEX             | INS_Flags_IsDstSrcSrcAVXInstruction)                                                                                                                                  // Return Minimum/Maximum scalar single